What happened
Base Labs, a research group spun up by Baseten earlier this year, has launched an open-weight AI safety partnership with Hugging Face and Goodfire.
Under the partnership, Base Labs will develop and publish methods for training and monitoring open models.
Why it matters
Open-weight models are widely distributed and can be inspected or modified by anyone who downloads them, so safety work aimed at them has to be published rather than kept internal to be useful.
By bringing together a research group with two named collaborators, the effort signals that open-model safety is being treated as shared infrastructure work rather than a single lab's private project.
Key facts
Base Labs is a research group that Baseten spun up earlier this year.
The partnership is with Hugging Face and Goodfire.
Base Labs will develop and publish methods for training and monitoring open models.
What to watch next
Whether the partnership releases concrete training and monitoring methods, and how openly they are published.
How the collaboration between Base Labs, Hugging Face and Goodfire takes shape as the work progresses.
